Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) is scheduled to launch
in late 2017 and will carry the Advanced Topographic Laser Altimeter System
(ATLAS), which is a photon-counting laser altimeter and represents a new
approach to satellite determination of surface elevation. Given the new
technology of ATLAS, an airborne instrument, the Multiple Altimeter Beam
Experimental Lidar (MABEL), was developed to provide data needed for
satellite-algorithm development and ICESat-2 error analysis. MABEL was
deployed out of Fairbanks, Alaska, in July 2014 to provide a test dataset for
algorithm development in summer conditions with water-saturated snow and ice
surfaces. Here we compare MABEL lidar data to in situ observations in
Southeast Alaska to assess instrument performance in summer conditions and in
the presence of glacier surface melt ponds and a wet snowpack. Results
indicate the following: (1) based on MABEL and in situ data comparisons, the
ATLAS 90 m beam-spacing strategy will provide a valid assessment of
across-track slope that is consistent with shallow slopes
(< 1
Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) is a NASA mission
scheduled to launch in 2017. ICESat-2 is a follow-on mission to ICESat
(2003–2009) and will extend the time series of elevation-change measurements
aimed at estimating the contribution of polar ice sheets to eustatic sea
level rise. ICESat-2 will carry the Advanced Topographic Laser Altimeter
System (ATLAS), which uses a different surface detection strategy than the
instrument onboard ICESat. Abdalati et al. (2010) provide an early overview
of the ATLAS concept and overall design. While the measurement goals of ATLAS
remain as described in Abdalati et al. (2010), some of the details have
evolved (Markus et al., 2016). ATLAS is a six-beam, photon-counting laser
altimeter (Fig. 1). In a photon-counting system, single-photon sensitive
detectors are used to record arrival time of any detected photon. ATLAS will
use short (< 2 ns) 532 nm (green) wavelength laser pulses, with a
10 kHz repetition rate, which yields a
Schematic ICESat-2 and MABEL beam geometry (dashed lines) and
reference ground tracks (grey lines along ice-sheet surface). ICESat-2 beam
pairs (separated by
Given this new approach to satellite surface elevation measurement, an
airborne instrument, the Multiple Altimeter Beam Experimental Lidar (MABEL),
was developed to (1) enable the development of ICESat-2 geophysical
algorithms prior to launch and (2) enable ICESat-2 error analysis. MABEL
(discussed in detail in McGill et al., 2013) is a multibeam, photon-counting
lidar, sampling at both 532 (green) and 1064 nm (near infrared) wavelengths
using short (
Following engineering test flights in 2010 and 2011, MABEL was deployed to Greenland (April 2012) and Alaska (July 2014) to collect data, including from glacier targets, and to assess elements of the resulting data that may vary seasonally. The Greenland 2012 campaign sampled winter-like conditions, while the Alaska 2014 campaign was timed to collect data during the summer melt season, which is characterized by open crevasses and surface melt ponds. In winter, increased albedo, reduced ice-sheet surface roughness, and reduced solar background and backscatter in the atmosphere all lead to an increased signal-to-noise ratio and an increase in photon-retrieval density (i.e., the number of, and temporal distribution of photons transmitted and recorded by the lidar). In general, with increased photon-retrieval density, we expect better surface measurement precision. In the extreme case, the photon-retrieval density may be sufficiently high that the instrument receiver does not have the time required to process the incoming photon information before receiving more. This effect is referred to as “instrument dead time” and can produce a positive surface elevation bias. In summer, reduced albedo, increased ice-sheet surface roughness, and increased solar background leads to a decrease in photon-retrieval density and signal-to-noise ratios, compromising measurement precision. The Alaska 2014 campaign also aimed to investigate how light at 532 and 1064 nm wavelengths interacts with the surface in melting conditions, and how this may affect the statistics of the 532 nm signal photons and overall elevation accuracy.
Here, we compare in situ measurements with MABEL airborne lidar data on the
Bagley (16 July 2014; 60.5
Map of the Multiple Altimeter Beam Experimental Lidar (MABEL)
flights used in this analysis from the July 2014 field campaign, which was
based out of Fort Wainwright, Fairbanks, Alaska.
MABEL data (Level 2A, release 9) for the Alaska 2014 campaign (Fig. 2) are
available from the NASA ICESat-2 website
(
The MABEL laser pulse repetition rate is variable (5 to 25 kHz) and was
5 kHz for the data presented here. At this nominal altitude and repetition
rate, and at an aircraft speed of
MABEL beams are arranged approximately linearly, perpendicular to the
direction of flight, with the 1064 nm beams leading the 523 nm beams by
Relative to one another, the MABEL beams have non-uniform average transmit energy. While all beams originate from a single 1064 nm laser source, each beam follows a unique optical path through the instrument once split from the source beam. Several individual beams maintain the fundamental 1064 nm wavelength of the source, while others are split off of a beam that is frequency-doubled to 532 nm (McGill et al., 2013). Owing to the frequency-doubling process and the non-uniform optical paths (fiber lengths) through the instrument, the 1064 and 523 nm transmit-pulse energies are generally not equal. During the 2014 Alaska campaign, there were fifteen 532 nm beams and six 1064 nm beams.
Our analysis used relatively high-energy beams. For analyses intended to
mimic the 90 m spacing of the ATLAS beam geometry, two 1064 nm beams were
chosen based on their across-track ground separation and along-track
signal-photon density: beams 43 (center of the array) and 48
(
Because of the different optical paths each beam takes through the instrument, each MABEL beam has a unique range bias (McGill et al., 2013). Prior to Level 2A data processing, MABEL ranges are corrected for these channel-specific optical path lengths using a calibration derived from data recorded during aircraft pitch and roll maneuvers performed over stretches of open ocean. We assume that this calibration mitigates the larger channel biases, including those associated with errors in pointing. However, other smaller-scale channel biases may still exist; these smaller-scale channel bias corrections were on the order of decimeters. Much of the analysis performed here, such as evaluation of local surface slope, did not require absolute range accuracy. Therefore, the individual beams were generally only calibrated with respect to one another based on data collected over the nearest flat surface (e.g., open water). These calibrations were made relative to the beam closest to the center of the array.
For the 2014 Alaska campaign, a camera was integrated with MABEL and was
successful for over 40 % of the campaign's duration. The images were
typically used to visually confirm the type of surface being overflown by
MABEL (e.g., ice, open water, sea ice, or melt ponds) or to confirm the
presence or absence of clouds. These images are also available on the
ICESat-2 website. The MABEL camera (Sony Nex7, with a 55 to 220 mm,
Data from the Landsat 8 Operational Land Imager (OLI) of the Bagley Icefield (Fig. 2b) were used as an independent assessment of the depths of melt ponds surveyed by MABEL. We applied spectrally based depth retrieval models to Landsat 8 imagery (Moussavi et al., 2016; Moussavi, 2015; Pope et al., 2016), which were calibrated based on data from supraglacial lakes in Greenland. We assessed the performance of OLI's coastal blue, blue, green, red, and panchromatic channels in retrieving supraglacial lake depths. Ultimately, the models establish a relationship between Landsat 8 top-of-atmosphere (TOA) comparing pre-drainage spectral reflectance values over the lakes with a post-drainage digital elevation model (DEM), derived from WorldView-2 imagery acquired from the Polar Geospatial Center at the University of Minnesota, using image-processing software (ERDAS). Our analysis indicated that, for shallow lakes (depth < 5 m), red and panchromatic band data are most suitable for supraglacial bathymetry. Because of the relatively small size of the lakes in our study area, we chose the panchromatic channel for the better spatial resolution.
A second WorldView-2-derived DEM was used near the terminus of the Lower Taku Glacier (Fig. 2c) to assess surface elevations derived from MABEL signal photons in steep and crevassed terrain. The DEM, created by the Polar Geospatial Center at the University of Minnesota, was extracted from high-resolution along-track stereo WorldView-2 imagery processed with NASA's open-source Ames Stereo Pipeline software (Moratto et al., 2010). The WorldView-2 images were collected on 6 June 2014, while the MABEL data were collected on 16 July 2014, and thus separated by 40 days. As part of an unrelated project, GPS data were continuously collected at six sites on the Lower Taku Glacier throughout the summer, using a Trimble NetR9 receiver; these data were used to tie the MABEL survey data to the WorldView-2 DEM. The data were processed kinematically using the Plate Boundary Observatory station AB50, located at the Mendenhall Glacier Visitor Center, approximately 20 km west of the survey area.
Previous studies (Brunt et al., 2013, 2014) have demonstrated that MABEL precisely characterizes the ice-sheet surface when comparing MABEL-derived slope on 90 m across-track length scales with those based on both Airborne Topographic Mapper (ATM; Krabill et al., 2002) and Laser Vegetation Imaging Sensor (LVIS, more recently referred to as Land Vegetation Ice Sensor; Blair et al., 1999).
We conducted a GPS survey on the Juneau Icefield (Fig. 2c) to determine the
length scale at which a ground-based local slope assessment on a flat surface
(< 1
GPS survey on the Juneau Icefield. Ground tracks for MABEL beams
43 and 48, from the 31 July 2014 flight, are indicated (red lines). GPS
survey points of the nodes of concentric, equilateral triangles, with side
lengths of 5, 10, 25, 50, 75, 90, 125, and 150 m, are indicated (black
points). Also indicated are the intersections of the MABEL flight lines with
the GPS survey grid (blue solid points), which were used to calculate MABEL
surface gradients (
MABEL-based surface gradients
We compared the MABEL-derived slopes to the slopes from each of the
concentric GPS triangles and the slope based on the GPS survey sites that
were closest to the nodes that defined the MABEL surface. Specifically, we
created a surface gradient comparison (SGC) parameter for each of the
GPS-derived triangles (
For illustrative purposes, we produced histograms of the MABEL surface return for the beams used in our analyses (Fig. 4; beams 5, 43, 48, and 50) from 3000 m of along-track data over a stretch of open ocean. We calibrated the beam elevations to one another to remove the unique beam elevation biases; relative bias corrections ranged from 0.03 to 0.73 m. We then detrended the surface elevations based on a linear fit to the signal photons to remove any elevation differences associated with wind stress or the relatively small effects of ocean dynamic topography and geoid undulation. The detrending of each beam takes into account all of these effects; this correction ranged from 0.11 to 0.29 m over the 3000 m of along-track data used for this analysis. We then produced histograms using a 0.01 m vertical bin size. We determined the full width at half maximum (FWHM) for each of the beams, which ranged from 0.19 m in beam 5 (532 nm) to 0.31 m in beam 43 (1064 nm). From Fig. 4, the relative differences in the signal strengths of the individual beams are evident from the non-uniform amplitudes of the photon-count distribution.
Histograms of the signal return for the MABEL beams used in this analysis (5, 43, 48, and 50). Plotted are ocean surface-return photon counts (per 0.01 m vertical bins) over a 3 km along-track distance against elevation (m). The elevations are calibrated to one another and detrended. The full width at half maximum (FWHM) for each histogram is indicated in the legend. The secondary return 0.75 m below the main signal return, which is more evident in the 1064 nm beams, is due to unintended secondary pulses from the MABEL laser that occur under some operational conditions; this was removed for FWHM analysis.
The MABEL signal often has a primary surface return and a second, weaker return approximately 0.5 to 1.5 m below the surface. This is due to unintended secondary pulses from the MABEL laser that occur under some operational conditions. The exact conditions for after-pulsing are not completely understood, but are most likely the result of temperature drifts in the fundamental laser system. These occur due to changing environmental conditions within the instrument pod in the aircraft, and/or changes in efficiency of the coolant system. The cooling system relies upon passive external fins exposed to ambient cold conditions at altitude, and these conditions (temperature, airflow) change during flight. The secondary laser pulses are primarily seen in the 1064 nm returns and are minimized when the 1064 nm source is frequency-doubled to generate 532 nm beams. This second pulse can affect statistics associated with MABEL results and was therefore manually removed. This secondary pulse is evident in the open-ocean data example at approximately 0.75 m below the main surface return (Fig. 4).
Given nearly uniform surface conditions, along-track signal-photon density for each beam varied within and between flights based on parameters such as weather conditions, time of day, and sun-incidence angle. The signal-photon densities on the Juneau and Bagley icefields, for each beam considered here, are given in Table 1. These densities are reported based on 0.70 m along-track length scales for direct comparison with previous results (Brunt et al., 2014), to mimic the ATLAS sampling interval (one laser shot every 0.70 m). MABEL along-track signal-photon densities for the July 2014 Alaska campaign were lower than those reported during the April 2012 Greenland campaign by Brunt et al. (2014). They reported 3.4 and 3.9 signal photons per 0.70 m for beams 5 and 6 (532 nm), respectively, over the Greenland Ice Sheet; the highest counts of signal photons per 0.70 m were 1.8 and 3.7 for 532 and 1064 nm channels, respectively (Table 1). Some of this variation may have been related to seasonal differences in surface reflectivity between the two campaigns, which include parameters such as the freshness of the most recent snowfall, the dust content of the surface, the presence (or absence) of surface melt and ponds, and the presence (or absence) of snow bridges that cover crevasses. Some variation may also have been related to instrumentation issues, such as cleanliness of the elements in the optics.
MABEL along-track signal photon densities over the open ocean and the Juneau and Bagley icefields.
The MABEL signal-photon densities (Table 1) are generally lower than those
expected for ATLAS. Under similar conditions as the 2014 MABEL summer
campaign and based on performance models, we expect the strong beams of ATLAS
to record 7.6 signal photons every shot (or 0.70 m along track) over ice
sheets and 0.5 to 1.8 signal photons every shot over the open ocean,
dependent upon the state of the wind (A. Martino, personal communication,
2016). We note that, for the Alaskan icefields, the expected number of signal
photons based on the performance model is probably too high, as the model
uses an albedo of 0.9, which is more appropriate for ice with fresh snow or
the interior of Antarctica than for icefields in Alaska in summer. Relative
to the performance model, at best (i.e., using data from beam 50) the MABEL
data used in this analysis suggest that the signal-photon densities were
We compared MABEL elevations to those based on the Juneau Icefield GPS array,
interpolated to the MABEL/GPS points of intersection (Fig. 3, blue solid
points). The mean offset, or bias, for the three points of intersection was
3.2
We assessed the surface precision of MABEL data (i.e., the spread of the
MABEL data point cloud about a known surface, or the standard deviation of
the mean difference between MABEL and a known surface elevation; Hodgson and
Bresnahan, 2004) over the flat stretch of open ocean used in the analysis of
Fig. 4. For approximately 3000 m of along-track open water, the
surface-precision estimates for the strong 532 and 1064 nm beams, based on
standard deviations of the mean differences from the detrended surface, were
We examined MABEL data from the Bagley and Juneau icefields and from the Lower Taku Glacier to determine how well photon-counting laser altimeters would capture surface detail on relatively short length scales (less than 1 km), such as crevasses and melt ponds.
Analysis of data from individual beams over the Bagley Icefield indicates
that MABEL can capture surface detail of crevasse fields. Figure 5a shows
stitched MABEL images of one set of crevasses on the Bagley Icefield; Fig. 5b
shows MABEL signal and background photons for a 1200 m range that includes
the glacier surface; and Fig. 5c shows MABEL signal photons, indicating
returns from both the glacier surface and the bottoms of a series of
crevasses. The along-track slope of this crevasse field, between 140.60 and
140.58
MABEL camera and photon data over a heavily crevassed section of
the Bagley Icefield, from the 16 July 2014 flight.
Similarly, analysis of the individual beams in a different area of the Bagley
Icefield indicated that MABEL can determine the location of melt ponds.
Figure 6a shows stitched MABEL images from crevasse and melt-pond fields on
the Bagley Icefield; Fig. 6b shows MABEL signal and background photons for a
1200 m range window that includes the glacier surface; Fig. 6c shows both
signal and background photon-count densities (per 125 shots, or
MABEL camera and photon data over crevasse and melt-pond fields on
the Bagley Icefield, from the 16 July 2014 flight.
Surface return and histogram of the signal return for MABEL beams
5 (532 nm) and 50 (1064 nm) over the
Analysis of data from individual beams near the terminus of the Lower Taku
Glacier (Fig. 8) demonstrates MABEL performance in regions with steeper
slopes. The slope in this region is 4
MABEL data over crevasse fields on the Lower Taku Glacier, from the
16 July 2014 flight.
MABEL-derived surface elevations over the Lower Taku Glacier were compared to elevations from the WorldView-2-derived DEM (Fig. 8b), which had 2 m horizontal resolution. Figure 8c is one of the images used to create the DEM shown in Fig. 8d. The MABEL data were collected 40 days after the WorldView-2 images were acquired. GPS data from the Lower Taku Glacier were used to determine mean ice-flow velocities to tie the two datasets together. Specifically, the MABEL ground tracks were migrated up ice flow, using the northing and easting components of the mean velocities derived from the GPS data, to more accurately compare MABEL surface elevations to those derived from the earlier WorldView-2 imagery. An elevation was then extracted from the WorldView-2 DEM for each migrated MABEL data point.
Mean ice-flow velocities varied substantially for the sites on the Lower Taku
Glacier (Fig. 8c). A mean ice-flow velocity of 0.2 m day
Using Eq. (1), we compared the MABEL-derived surface-gradient comparison
(SGC) parameters to those based on the Juneau Icefield GPS array (Fig. 9).
The MABEL-derived SGC parameters were consistent with GPS-derived SGC
parameters over length scales ranging from 50 m (just over half of the ATLAS
beam spacing) to 150 m (just under twice the ATLAS beam spacing). The SGCs
for 50 to 150 m spatial scales were less than 0.5
A surface-gradient comparison between a MABEL-derived surface
(blue points in Fig. 3) and a series of GPS-derived surfaces, based on
concentric equilateral triangles (black points here and in Fig. 3) and a
surface based on the GPS survey sites that were closest to the nodes that
defined the MABEL surface (blue point here and blue open circles in Fig. 3).
The
The high-resolution WorldView-2 DEM also provided a means of assessing
MABEL-derived across-track slopes in steeper glacial settings. Using a method
similar to that of Brunt et al. (2014), we calculated a
MABEL and DEM surfaces and slopes for a small stretch (see box in
Fig. 8b) on the Lower Taku Glacier.
As noted above, there are some significant differences between MABEL and ATLAS depicted in Fig. 1 (e.g., number of beams, beam pattern, and altitude) and described elsewhere in this paper (e.g., footprint size, along track footprint spacing, and wavelengths). In order to relate the predicted performance of ATLAS with the measured performance of MABEL, some common metric is necessary that accounts for as many of the differences as is practicable. The signal-photon density is a metric to relate the radiometry of the two instruments. Given that the signal-photon density is generally less than that predicted for ATLAS, for a given background rate, the surface should be more easily distinguished in ATLAS data. While in theory one could use the framework developed for predicting ATLAS radiometric characteristics to make similar predictions for MABEL and therefore use MABEL data to evaluate that framework, the efficiency or radiometric throughput of MABEL has not been characterized well enough to do so. Flight data (Brunt et al., 2014) show that, for a given campaign, the measured signal-photon density of MABEL changes by tens of a percent over a relatively uniform ice sheet interior. Similar changes are measured for the background rate, after consideration for sun angle is taken into account. As such, the analysis presented here cannot be used to quantitatively assess the likelihood that ATLAS will meet its measurement requirements (or the mission science objectives). What we can say is that if the ATLAS signal-photon density and signal-to-noise ratios are within 30 % of its measurement requirements (and thus mimics the MABEL performance documented in this study), ATLAS can be used to measure surface slopes over both relatively flat ice-sheet interior conditions and steeper glaciers, such as the Lower Taku Glacier, and identify melt ponds. If ATLAS fully meets its measurement requirements, we expect that the definition of small-scale surface features such as crevasses and melt ponds will be correspondingly improved.
The result of this analysis indicates that the MABEL-derived local slope
assessment, on a relatively flat glacial surface and on a 90 m across-track
length scale, is consistent with in situ slope assessments made at spatial
scales ranging from 50 to 150 m. For a planar surface where slope is less
than 1
Based on our comparison with a WorldView-2-derived DEM of the Lower Taku
Glacier, MABEL can also provide valid estimates of across-track slope, even
in steeper terrain. Once migrated for GPS-derived ice-flow displacements, the
southern part of the MABEL-derived surface elevations is in good agreement
with the DEM data, and the slope comparison between MABEL-derived and
DEM-derived across-track slopes had a mean residual of 0.25
Figures 5c and 6d suggest that the dense along-track sampling of MABEL is sufficient to capture surface detail, including melt-pond information, from a single, static beam in regions of low slope, consistent with that of an ice-sheet interior. Based on the continuous nature of the surface return through the crevasse field, especially in the 1064 nm beam (50) in Fig. 5c, we conclude that MABEL frequently retrieves a signal from the bottom of crevasses. Further, Fig. 8b indicates that MABEL continues to provide surface detail in regions of steeper slope, including the retrieval of the steep slopes of the crevasse walls (e.g., Figs. 5c and 6d).
As previously noted, MABEL data used in this analysis had signal-photon
densities that are
The crevasse characterization we performed on the Bagley Icefield is qualitatively confirmed using the camera imagery (Fig. 5a). However, it should be noted that we have no means of quantitatively assessing the accuracy of MABEL-derived crevasse depths. Crevasses on an ice-sheet surface have an influence on albedo (Pfeffer and Bretherton, 1987). This variation in reflectance is evident in Figs. 5b, 6b, and c, where MABEL background photon counts, and the signal-to-noise ratios, change significantly. Changes in MABEL background photon densities have also been used to detect leads in sea ice (Kwok et al., 2014; Farrell et al., 2015). From Fig. 6c we note that the overall background photon counts decrease significantly over the eastern region of this plot, which is characterized by crevasses; however, this change is non-uniform. Background photon counts drop steadily to nearly zero over the two melt ponds surveyed along this transect.
Penetration of 532 nm wavelength light into the surface, be it a melt pond
or snow, is an ongoing area of research for ICESat-2 algorithm development.
MABEL geolocation uncertainty, and the fact that the 1064 and 532 nm beams
do not have coincident footprints for more direct comparison (as the 1064 nm
beams lead the 532 nm beams by
Analysis of MABEL data over small melt ponds on the Bagley Icefield in Alaska provided a preliminary assessment of how green-wavelength photon-counting systems will interact with water on an ice surface. Based on the signal-photon elevations in Fig. 6d, and the histogram of the signal photons in Fig. 7, the total spread of the signal photons, at a wavelength of 532 nm, is approximately 1.5 to 2 m. Further, analysis of Landsat 8 and WorldView-2 imagery confirm that the melt ponds in this region are approximately 2 m deep. These results suggest that, while there is not a distinct signal return from a melt-pond bottom, the 532 nm MABEL beam may be sampling the entire melt-pond water column. The 1064 nm MABEL beam shows evidence of a secondary return 1.5 m below the main signal return, due to unintended secondary pulses from the MABEL laser that occur under some operational conditions, and is likely not due to melt-pond bottom returns.
Based on the surface characterization results of MABEL data from the Juneau and Bagley icefields, and the dense, six-beam sampling strategy of ATLAS, we are confident that ICESat-2 will contribute significantly to glacier studies at local and regional scales and in polar and mid-latitudes. While previous studies using satellite laser altimetry have investigated the vertical dimension of rifts in the ice sheet (e.g., Fricker et al., 2005), those studies have been limited to major ice-shelf rift systems, as opposed to smaller-scale crevasses. The 0.70 m along-track sampling density of each individual ATLAS beam is well suited for similar vertical dimension studies, but at finer length scales, such as those associated with alpine glacier crevasse fields.
Knowledge of local slope and local surface character are required to
accurately determine ice-sheet surface-elevation change. The ATLAS beam
geometry includes pairs of beams separated at 90 m across track to enable
the determination of local slope in one pass, and therefore to enable the
determination of ice-sheet surface-elevation change in just two passes.
Based on the analysis of MABEL, ground-based GPS data, and the resultant
surface gradient comparison (SGC), we conclude that the ATLAS 90 m
beam-spacing strategy will provide a valid assessment of local slope that is
consistent with the slope of an ice-sheet interior (< 1
The MABEL 2014 Alaska campaign was timed to collect data during the summer melt season to specifically investigate how 532 nm wavelength laser light interacts with a melting snow surface. Results from MABEL, and confirmed through analysis of Landsat 8 imagery, suggest that 532 nm wavelength light is likely reflecting from the surface and subsurface of the 2 m deep supraglacial melt ponds on the Bagley Icefield. This is an ongoing area of research for ATLAS and ICESat-2 algorithm development.
MABEL lidar data and camera imagery are publicly available on the NASA
ICESat-2 data page
(
Funding for this project was through the NASA ICESat-2 Project Science Office. Funding for J. M. Amundson was provided by NSF-PLR 1303895. We acknowledge the considerable efforts of the Project, Science, and Instrument teams of NASA's ICESat-2 and MABEL missions. We thank the following people: Eugenia De Marco (ASRC Aerospace Corp., NASA/GSFC) and Dan Reed (Sigma Space Corp., NASA/GSFC) for MABEL instrument support; Scott Luthcke (NASA/GSFC), David Hancock (NASA/WFF), and Jeff Lee (NASA/WFF) for MABEL data calibration; Scott McGee and Ya' Shonti Bridgers (JIRP) for GPS field data collection and data processing support; and NASA/AFRC (specifically ER-2 pilots Tim Williams and Denis Steele) for Alaska airborne support. WorldView imagery was provided by the Polar Geospatial Center at the University of Minnesota, which is supported by NSF-PLR 1043681. GPS receivers for the survey of the terminus of the Lower Taku Glacier were provided by UNAVCO. GPS receivers for the JIRP survey were provided by Werner Stempfhuber of the Beuth University of Applied Sciences. And, finally, we thank two anonymous reviewers for their highly constructive suggestions. Edited by: A. Kääb Reviewed by: two anonymous referees